Modern cap replacements for audio crossover

[Traceless]

Hey everyone,

a capacitor in one of my speakers' audio crossover went bad. Unfortunately the capacitors used in the crossover are not manufactured any more. I probably could just drop in a regular electrolytic with the same capacitance and volrage rating, however I imagine that the audio crossover caps may have special characteristics e.g. lower ESR at higher frequencies, stricter tolarances etc. Can anybody recommend a suitable modern* replacement for the the "Tesla TF202 4.7µF 50V" and the "Ton Frequenz 10µF 100V" shown on the picture below?

*I'd rather not use new old stock which will probably be either dead on arrival or won't last very long.

[bdunham7]

Any bipolar electrolytic with reasonably low tolerances should be fine.  Note that bipolar electrolytics are much less common than polar ones.  Some like replacing these with polypropylene film capacitors for lower distortion, but there will be a slight change in filter response as well because the PP have much lower ESR. PP will also be physically much larger.

Examples:

https://www.mouser.com/ProductDetail/Cornell-Dubilier-CDE/475BPA050M?qs=sGAEpiMZZMsKEdP9slC0YVPN%2FqWokS48XTZdzwk6rFQ%3D
https://www.mouser.com/ProductDetail/Cornell-Dubilier-CDE/106BPA050M?qs=sGAEpiMZZMsKEdP9slC0YT%2FRKTM8hWCgGd6Bn%252B%2FUuNg%3D

Although the 20% tolerance doesn't seem great, they're probably good enough.  Or, for PP:

https://www.parts-express.com/speaker-components/crossover-components/crossover-capacitors/metalized-polypropylene-crossover-capacitors/brand/Dayton-Audio

[Twoflower]

Check your preferred electronic shop (not Ebay) for "10µF bipolar" and "4,7µF bipolar" as replacement. They cost less than 2€ each (e.g. Reichelt). You can use higher voltage ratings without a problem as long as the dimensions are acceptable (Reichelt has only 4.7µF 63V or 100V).

As alternative you might consider Foil-Caps they are also not polarized. For example MKS4-100 10µ, MKS4-50 4,7µ But they might not fit the holes in the PCB as the leads are not axial and cost more (still under 2€ each). They supposed to last longer if operated within specification.

[Traceless]

Thanks bdunham7 and Twoflower for the response. I did not realize that these were bipolar, but instead assumed that the indent on the left side would mark the cathode, similar to these ones here.

[abdulbadii]

its kind spread out on market, not have to be Tesla aslong as non polarized low ESR quality
https://roundupreviews.com/us/audio-crossover-capacitors

[CaptDon]

You will find almost any of recent manufactured non-polar caps to have better specs than the originals. Some of the older caps had wide tolerance on the exact capacitance and higher ESR than anything made today. As for an ESR of .1 ohm vs even something as high as 1 ohm it is doubtful you could tell the difference in a blind listening test. If you look at the true impedance across a speakers design frequency operating range an 8 ohm speaker could be anything from its DCR of perhaps 6 ohms to as high as 40 ohms. If you take the worst case example where the speaker impedance is low and you compare .1 ohm ESR vs. 1 ohm ESR capacitors the difference in produced Sound Pressure Level is barely above the minimum 'difference in level' detectable at steady state tone. Far less detectable in a complex signal such as music.
 

[Kleinstein]

With only 4.7 and 10 µF, I would use film capacitors. They are available relatively well, have usually low tolerance, low ESR and they essentially don't age very much. So even an old film cap is OK.

I one wants PP caps, one could also look at motor run capacitors. They are bulky, but the capacitors itself are good and available also to large values (e.g. 50 µF). Nomally MKS (polyester / mylar) film capacitors should be good enough and they are smaller.

[Traceless]

Thanks again to everyone here for their feedback I ordered replacement caps which should arrive next week. For my initial diagnosis I did some quick and dirty measurements of the caps with my multimeter. The Ton-Frequenz looked okay (11.7µF), however the Tesla seemed to be way out of spec (11.28µF) so I assumed it was faulty. Now that I got the audio crossovers out of the speaker housing and did some proper measurements of the caps on my bench I found that the Tesla measures 4.795µF, ESR 0.28 Ohms at 12 kHz. At lower frequencies things look quite different though, at 100Hz it measures 7.7µF with an ESR of 48.7 Ohms, at 300 Hz 6.56µF, ESR 16.33 Ohms, at 1kHz 5.67 µF with an ESR of 4.5 Ohms. In contrast the Ton-Frequenz cap stays a lot closer to its nominal value even at low frequencies and measures 10.88µF with an ESR of 4.42 Ohms at 100Hz. At 12 kHz the Ton-Frequenz measures 10.67µF with an ESR of 0.0475 Ohms. Does anyone know which frequency the Tesla caps are rated for? I just wonder, if it is possible that the Tesla cap is made for high frequency purposes only, and therefore just appears to be bad at lower frequencies? Unfortunately I didn't find any spec sheet for the Teslas, but maybe someone here has experience with these?

[CaptDon]

I don't have any specs but one observation. Depending on how the ESR meter does its calculations they can be way off from the truth. We have at least three brands of ESR meters in the lab, one being a General Radio 'lab type' L/C/R meter and between the three units the ESR readings vary wildly but the capacitance seems closer across the bunch. Smaller (1uf and under) capacitors tend to have the largest ESR disagreement between units and large capacitors (1000uf and up) tend to really be a crap shoot when tested at higher frequencies. The hardest to test in our lab was the 1700uf@2500vdc units (around 60 pounds each) used in locomotive link voltage filtering. We needed extremely low ESR and with the physical size of the units it was hard to have low self inductance which would show up as 'ESR'. It is true the self inductance was an 'impedance' and every impedance has some 'R' associated with it. The 'true' ESR was in milliohms but when tested at high frequencies the self inductance made the ESR look like tens of ohms. This was a problem for square wave switching currents in the thousands of amps!! B.T.W., There are no less than 12 of these monster caps paralleled across the link voltage and 4500 horsepower of electrical energy being delivered to the link at full throttle. When one of these caps fails shorted at full throttle they have been known to blow the doors of the auxilliary cab off the locomotive followed with an oil fire.

[Traceless]

We have at least three brands of ESR meters in the lab, one being a General Radio 'lab type' L/C/R meter and between the three units the ESR readings vary wildly but the capacitance seems closer across the bunch. Smaller (1uf and under) capacitors tend to have the largest ESR disagreement between units and large capacitors (1000uf and up) tend to really be a crap shoot when tested at higher frequencies.

@CaptDon: That is an interesting observation, thank you for sharing that. I did not expect LCR meters to deliver wildly different readings, unless they are set to the wrong measurement frequency, not properly calibrated or broken. Very recently I upgraded from my cheap component tester (which was more of a guesstimometer, when it came to ESR measurements) to a proper LCR meter. Before pulling the trigger on my meter I read through this forum and watched some reviews of the usual suspects like the East Tester and the DER EE DE-5000. For instance in Defpom's review (link below) both devices seemed to more or less agree. Unfortunately I don't have a second LCR-Meter for a sanity check, but specs-wise the MCR 5200 I chose should be pretty accurate (you can see it in action in for instance Tony's review). I did some measurements with my unit on various newly bought electrolytics and all readings seemed very reasonable. Only if I dial in the wrong frequency, measurement results are often way out of spec (which is to be expected).

Depending on how the ESR meter does its calculations they can be way off from the truth.

Chapter 4 of the meter's manual gives some information about how it operates and what accuracy can be expected. Given that the 50V Tesla capacitor I desoldered from the audio crossover has a nominal capacitance of 4.7µF and measures 4.795µF at 12 kHz my best guess was that the cap is actually designed for that frequency. I guess once the replacement caps have arrived I could repeat the measurements I did on the Tesla cap to rule out issues with my meter.

[Kleinstein]

Electrolytic capacitors have a rather high amount of dielectric absorbtion (DA). One consequence of this is the the capacitance will go up towards lower frequency. Something like 10% more with 1/10 the frequency would be perfectly normal and a bit more could be seen if the capacitos has high levels of DA.
For a comparison one could do at least a crude test on the DA.

The change in capacitance seen for the Tesla cap looks rather high.

[bdunham7]

I did not expect LCR meters to deliver wildly different readings, unless they are set to the wrong measurement frequency, not properly calibrated or broken.

An audio crossover capacitor should typically test reasonably at 1kHz and 10kHz, with the possible exception of very large ones.  When you start seeing results that far off, you need to measure some other parameters, like DC leakage.  On a good PP film cap, I would expect all non-broken capacitance meters of any theory of operation to give similar results at any non-absurd test frequency.  On a good electrolytic, I might expect some variance, especially with frequency.  On a leaky, degraded or otherwise crappy electrolytic, I'd expect widely varying results like you have.  I wouldn't blame the meter, although I prefer to use a meter that gives you DCR, impedance and phase angle as separate readings.

[Traceless]

@Kleinstein: When we're talking about a "crude test on DA" I guess that means attaching my DMM to the cap charging it up with my bench-PSU discharging it and monitor if it builds up any noteworthy voltage afterwards?

An audio crossover capacitor should typically test reasonably at 1kHz and 10kHz, with the possible exception of very large ones. 

Great, I'll keep that rule of thumb in mind when testing audio crossover caps. Sure I can take a look at leakage current. Since both you and Kleinstein already mentioned that the Tesla cap is very likely too far out to still be healthy and based on the fact that the (2-way) speaker only produced muffled sound on the mid-range/woofer-unit but the tweeter was still pretty much okay - the initial assessment "broken" cap was likely correct afterall.

I wouldn't blame the meter, although I prefer to use a meter that gives you DCR, impedance and phase angle as separate readings.

IIRC you can get all these readings on the meter if you toggle the display mode, I just got it recently and haven't had much chance to use it yet.

Thanks again to everyone here for the helpful feedback!

[Kleinstein]

The crude DA test is charging, some 5 s of discharge and than after some 3-15 min check the voltage that has build up. The US tsndard want a relatively longe waiting time, but a shorter time is less critical to leakage and may be easier to measure. With a normal DMM with 10 M one has to be relatively fast and disconnect the meter in between. With a good high Z meter one could keep the meter connected.
With an electrolytic cap. Some 10% of voltage recovery would be not unusual.

[Traceless]

I replaced both caps on both audio crossovers this WE. The speakers work again. The capacitance on the replacement 4.7µF caps was also a little high but still within spec - according to my DMM (~5.41µF). On my LCR meter the replacement caps measured 4.769 µF, ESR 1,7 Ohm at 1kHz, 4.499µF, ESR 0.28 Ohm at 10 kHz.

@Kleinstein I tested dielectric absorption on one of the old Teslas, hooked it up to a DC power source at 50V, disconnected power, then discharged the cap for 5 seconds. After 10 minutes (the DMM was disconnected during the waiting period) I measured a voltage very close to 1V.

@bdunham7 I also measured leakage current of the presumably defective Tesla caps. When fully discharged I measure around 4mA right after turning on the DC power source. This value drops to 1-2 mA after a couple of seconds and below 1mA a bit later.

Semi-Off topic question: The old caps were glued to the board with some kind of black glue. Apparently this stuff turns soft when applying IPA, still it was a real hassle to get the caps off and involved IPA, knife, screwdriver and finally twisting the caps off with pliers. Has anyone encountered that black stuff and knows a better method to remove it?

[bdunham7]

@bdunham7 I also measured leakage current of the presumably defective Tesla caps. When fully discharged I measure around 4mA right after turning on the DC power source. This value drops to 1-2 mA after a couple of seconds and below 1mA a bit later.

That sounds very high for a 4.7uF capacitor.  Of course you need to test it both ways since it is bipolar. 

[Traceless]

That sounds very high for a 4.7uF capacitor.  Of course you need to test it both ways since it is bipolar. 

I just repeated the tests it behaves similar both ways. However after not being charged for about a week the cap behaved even worse. It ran even into the 20mA current limit I set for the test. Here is the recording of the first run. If I discharge the cap and the re-run the same experiment the current stays a lot lower, as can be seen here.

[bdunham7]

I just repeated the tests it behaves similar both ways. However after not being charged for about a week the cap behaved even worse. It ran even into the 20mA current limit I set for the test. Here is the recording of the first run. If I discharge the cap and the re-run the same experiment the current stays a lot lower, as can be seen here.

Just as a comparison I have a 4uF bipolar from some 40-year old speakers and I set it up with 50VDC.  It was instantly under 1mA (faster than my meter could read) and under 5uA in a few seconds.  So your leakage probably accounts for the varying test values, but you'd have to use a different tool to really see what is going on--getting a graph of Z and theta vs frequency or using an octopus tester.  I wonder how badly that leakage would affect its value as a crossover capacitor where it never gets DC to reform it.

[Traceless]

Just as a comparison I have a 4uF bipolar from some 40-year old speakers and I set it up with 50VDC.  It was instantly under 1mA (faster than my meter could read) and under 5uA in a few seconds. 

I tested the 63V, 4,7µ caps for reference. They also almost charge up instantly and very briefly draw 0.015mA but then drop immediately to 0 mA as expected.

So your leakage probably accounts for the varying test values, but you'd have to use a different tool to really see what is going on--getting a graph of Z and theta vs frequency or using an octopus tester. 

I've never heard of an octopus tester, may I ask what kind of tool that is?

I wonder how badly that leakage would affect its value as a crossover capacitor where it never gets DC to reform it.

Well about as bad as can be seen in this video (I had pretty much the same symptoms*1, very different speakers though):
https://youtu.be/c4mw1X48Z9M?t=120

Edit: *1 - in the video it seems the high-pass is broken, for me it was the low-pass.

[bdunham7]

I've never heard of an octopus tester, may I ask what kind of tool that is?

AKA V-I curve tracer, component tester, etc.  Used with an oscilloscope in the XY mode.

https://www.jammarcade.net/simple-component-tester-a-k-a-octopus-curve-tracer/